With the widespread use of digital cameras, including digital cameras provided as part of a mobile phone or other hand held device, many camera users are able to take a large number of images. Although many users of such devices enjoy the convenience with which digital images may be obtained, having unsorted digital libraries of many images can make navigating the images very time consuming. Users sometimes manually “tag” each image with a descriptor or several descriptors that identities the image.
Some digital cameras may provide intelligent suggestions for suitable descriptors for new images based on a database of tags and corresponding image feature vectors, which numerically represent the features in an object image. Some databases of image descriptors use the description of images provided by a user to train the database of tags and image feature vectors by adding the image feature vector of new images and the corresponding tags into the database. However, these techniques can introduce inaccurate tags if a user adds inaccurate tags to an image.